Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×

Comment Re:Maybe Apple just has the better position? (Score 1) 250

Having read significant chunks of Apple's response: Either Apple is lying to the court about legal precedents or DOJ is lying to the court about those precedents, and it should be a simple matter to determine which one is the liar. And lying to the court should be cause for disbarment and for further sanctions.

FWIW.

Comment This speed limit is reckless (Score 5, Informative) 582

It is very well established in the civil engineering literature how to set the speed limit for safety. Outside of exceptional circumstances the best rule is to establish the speed limit at the 85th percentile of traffic speeds (i.e., so that 85% of the traffic is going less than that limit). Unfortunately, politicians think they know better than the established engineering. They think that their wishes trump the facts.

Those who set that speed limit are acting in reckless disregard for the safety of the public. As is that CS professor--and he should know better!

Comment RXVT (Score 1) 352

...and is customizable from the command line: I have a set of about three dozen color-customizations, one for each remote host I work on. My menu has "run an ssh remote session on this customization of RXVT" for each of those hosts. Can't do that with most of these new-fangled terminal emulators...

Comment Re:long methods and coupling (Score 2) 497

Two points: experimental, and cognitive
Experimental:
the evidence shows that the minimum bug-rate occurs for methods of somewhere between 300 and 500 lines. More than that and it really is too complicated to comprehend, most of the time. Less than that and you think "It's simple" and don't pay proper attention.
Cognitive: magic numbers
  • 2 - 3 is the number of levels of recursion you can keep in your head at one time. Nested logic structures and call-stacks should never require you to look more than 3 levels deep to understand them.
  • 6 - 7 is the number of distinct objects you can keep in your head at one time. Never require that the reader keep track of more than that. If you think you need to, it means you haven't properly thought out your problem yet.

Side-note: Names that are too long fall afoul of the same effect: they force you to concentrate on how to read the name, rather than what the object means. Complex names should only be used for rarely-touched global entities; local scratch-variable names should be shorter -- the extreme being single-letter "i" for loop counters:

for( int i=0 ; i<dingbat_max ; i++ ) {...

Comment Re:You're the problem (Score 3, Informative) 497

Donald Knuth has stated that he is in favor of these kinds of "structured GOTO's".

There is a theorem that in general, GOTO's can be replaced by WHILE's and flags. What is generally not mentioned is that in the general case, the size of the resulting code grows exponentially in terms of the size of the original code. ;-(

As it happens, we do not get exponential growth for the "break out of the loop" cases...

Comment Mickey Mouse copyright extenstions... (Score 4, Insightful) 183

It is not fair.

Neither is it fair that Disney stole Osamu Tezuka's Kimba for use in The Lion King.

The Constitutional requirement is: (1) to authors and inventors, (2) for a limited time, (3) in order to promote progress in the sciences and arts.

It is impossible that extending the copyright term for works of a fifty-year-dead author can encourage him to produce more work. Nor is the resulting term "limited" in either in mathematical or human terms. And the current Mickey Mouse "copyright owners" are certainly NOT that author nor inventor.

Comment Re:Haswell-EP Xeons (Score 1) 150

You've never run CFD,have you? ...and don't know its message-patterns, either. That's what the OP said he was doing (though he didn't say whether he was running someone else's "canned" code, or was compiling his own). Unless it's specifically compiled for Haswell (which is unlikely,but sad), the "canned" code will not take proper advantage of the Haswell AVX2 instruction-set.

Comment Haswell-EP Xeons (Score 2) 150

I would go with Haswell-EP Xeons -- probably 2697v3 (14 cores @ 2.6-3.6): a two-socket motherboard gives you 28 physical cores per board, for prices in the $12K range. Just one of these is quite a powerful system. If you can get by with a 2-node system, then 10GE interconnect is good enough (AND MUCH CHEAPER); for more nodes, you will need Infiniband (since 10GE does not scale well). The 4-node/IB cluster will be on the order of $60K, and will offer more performance than a $160K solution of a couple of years ago.

These will offer far better performance than the Opteron solution.

Can you compile your own application? If so, use the Intel compilers, and make sure you compile targeting the Haswell instruction set (-O3 -Xhost -march=corei7-avx2 -mtune=corei7-avx2 if I recall correctly): the full AVX2 Haswell instruction set is rather more powerful for your app than the predecessor "AVX" SandyBridge/IvyBridge instruction set, which is far more powerful than the previous Nehalem/Westmere SSE4.2 instruction-set, which is somewhat more powerful than a simple "-O3". If you can't compile on your own, try to make sure the vendor's executables target AVX2; the right compile-flags will double your performance over "-O3"...

Slashdot Top Deals

What this country needs is a dime that will buy a good five-cent bagel.

Working...